Clinical Evaluation of a Fully-automatic Segmentation Method for Longitudinal Brain Tumor Volumetry

Area of Science:

• Neuro-oncology
• Medical imaging analysis
• Artificial intelligence in medicine

Background:

• Accurate tumor volumetry is crucial for diagnosing and treating brain tumors.
• Longitudinal monitoring of tumor size and evolution aids in treatment assessment.
• Manual segmentation is time-consuming and subject to inter-rater variability.

Purpose of the Study:

• To evaluate the efficacy of the automated segmentation tool BraTumIA for longitudinal brain tumor volumetry.
• To compare BraTumIA's volume estimations with manual segmentation (ground truth).
• To assess BraTumIA's potential to replace manual segmentation in clinical follow-up.

Main Methods:

• Analysis of longitudinal Magnetic Resonance (MR) Imaging data from 14 glioblastoma patients.
• Inclusion of 64 MR acquisitions from pre-operative to 12-month follow-up.
• Comparison of BraTumIA's automated segmentation volumes against manual segmentation by two raters.

Main Results:

• Strong correlations (R=0.83-0.96) between BraTumIA and manual raters for contrast-enhancing (CET) and non-enhancing T2-hyperintense tumor compartments (NCE-T2).
• Inter-rater disagreement between BraTumIA and human raters was comparable to that between human raters.
• BraTumIA produced volumetric trend curves similar to those from manual segmentation.

Conclusions:

• BraTumIA demonstrates reliable performance in longitudinal brain tumor volumetry.
• Automated segmentation shows potential to substitute manual volumetric follow-up for CET and NCE-T2 tumor compartments.
• This automated approach could streamline patient monitoring and improve diagnostic accuracy.